Bug 853350 is going to add a fanotify-based component that will notify about a low storage situation. We need to observe this notification and avoid any non-certified app to write in indexedDB until the low device storage situation is solved.

Ben, can you dig in here?

Did you mean to unnom, minus, or +? Can't tell by comment 1 what the blocking call is.

I guess Johnny meant to tef+. IMHO this bug needs to block for the same reasons as for example bug 861920 and bug 861894.

This works for b2g18. Minimal changes, we're going to return a QuotaExceededError whenever we get into low disk mode. Depends on the patch in bug 853350. Also has a huge test.

That was an old patch. This is the right one.

Comment on attachment 741307 [details] [diff] [review] Patch for mozilla-b2g18 Review of attachment 741307 [details] [diff] [review]: ----------------------------------------------------------------- looks good, r=me ::: dom/indexedDB/OpenDatabaseHelper.cpp @@ +1753,5 @@ > + NS_WARNING("Refusing to create database because disk space is low!"); > + return NS_ERROR_DOM_INDEXEDDB_QUOTA_ERR; > + } > + } > + Nit: It seems there are only two callers of OpenDatabaseHelper::CreateDatabaseConnection() and the one in FileManager.cpp always calls it when the db file exists. So the check added here could just live in OpenDatabaseHelper::DoDatabaseWork() I think. But that's not a big deal.

Now we need to make sure that we have a sane way to do error conversions from SQLITE_FULL to QuotaError. This is the groundwork for a messy patch that changes a bunch of |NS_ENSURE_SUCCESS(rv, NS_ERROR_INDEXEDDB_UNKNOWN_ERR)|...

Comment on attachment 742312 [details] [diff] [review] Part 2: Simplify error compression (mozilla-b2g18) It's not clear that we need this for b2g really... Let's wait on the review.

FTR a "FreeSpaceWatcher" was created in Bug 818848 for the Webapps API. It was done in JS but I wonder if it's not possible to do it once (maybe in C++) and use it from everywhere it is needed.

The space watcher component created in Bug 818848 is using a polling mechanism to check for device space while an app is being installed. We decided not to use the same strategy for Bug 853350 as having a component continuously polling for device space would be bad for performance and battery consumption. We chose to use a real time notifications strategy based on fanotify. Check Bug 853350 for more details about it. Also, Bug 863596 will change the Webapps FreeSpaceWatcher component to observe notifications triggered by the fanotify based component.

What are the next steps here?

Bug 853350 needs to land. And then I need to write a trunk patch (probably a different bug though).

Merged to trunk.

Comment on attachment 747566 [details] [diff] [review] Patch for mozilla-central Review of attachment 747566 [details] [diff] [review]: ----------------------------------------------------------------- ::: dom/indexedDB/IndexedDatabaseManager.cpp @@ +119,5 @@ > + nsCOMPtr<nsIDiskSpaceWatcher> watcher = > + do_GetService(DISKSPACEWATCHER_CONTRACTID); > + if (watcher) { > + bool lowDiskSpaceMode; > + if (NS_SUCCEEDED(watcher->GetLowFreeSpace(&lowDiskSpaceMode))) { You probably noticed this, but the idl for nsIDiskSpaceWatcher changed. So you have to s/GetLowFreeSpace/GetIsDiskFull @@ +123,5 @@ > + if (NS_SUCCEEDED(watcher->GetLowFreeSpace(&lowDiskSpaceMode))) { > + sLowDiskSpaceMode = lowDiskSpaceMode ? 1 : 0; > + } > + else { > + NS_WARNING("GetLowFreeSpace failed!"); ditto

The same comments apply for the b2g18 patch

Updated for newer nsIDiskSpaceWatcher changes.

https://hg.mozilla.org/mozilla-central/rev/22a77fd7ab35 https://hg.mozilla.org/mozilla-central/rev/7a362c86108c https://hg.mozilla.org/mozilla-central/rev/462c163822b3 https://hg.mozilla.org/mozilla-central/rev/c26ceab09064 https://hg.mozilla.org/mozilla-central/rev/ee537efbe1e2 https://hg.mozilla.org/mozilla-central/rev/8db9762409f4 Hilarity.

Ben, I'm going to leave this uplift for you since I have no confidence that the b2g18 patch attached here includes the fixes it took to get this compiling and I have no time to go through it today myself.

Turned the test on for most platforms. https://hg.mozilla.org/integration/mozilla-inbound/rev/77b8505fdd33

lgtm on 1.01 with a partner build. Tested in a similar fashion to what was done with appcache here - https://bugzilla.mozilla.org/show_bug.cgi?id=861894#c60.